Design And Experimental Research On Automatic Detection System Of Fabric Defects

Fabric defect is an important factor affecting the quality of textiles, and defect detection is the essential link in textile production process. However, different kinds of defects can be generated in any stage of the process. Because of the fabric is flexible, the defect has been proved to be the most difficult to be automatic detected in the production process of textile.The theory research and equipment development of automatic detection of fabric defects at home and abroad has been quite comprehensive analyzed in this thesis. Based on this, the theoretical analysis and experimental study are carried out in the automatic defects detection system which based on computer vision.First of all, according to the technological requirements of defect detection in actual production and based on the existing cloth inspection machine, the overall structure of the detection system is optimized designed in this thesis. Line array CCD camera is selected to collect image. The matching of camera scanning rate and cloth transferring rate, illumination light source and the software design scheme of the detection system are presented in this thesis.Secondly, the causes and characteristics of noise which produced in the process of image acquisition are analyzed. Through the comparison of common filtering methods, the method of mathematical morphology is used to smooth the image. In order to reduce the loss of image detail information caused by image smoothing, Laplace operator is used to image sharpening.Thirdly, the Canny operator is used to detect the edge of the fabric image. According to problems existing to the classical Canny operator in handling the images in real-time, a improved adaptive threshold canny detection algorithm is presented in this thesis, which effectively improves the real-time performance of Canny operator.Finally, the automatic defect detection experiment system is designed based on the enterprise existing cloth inspection machine. Image acquisition of the system is realized by "computer + line array CCD camera scanning; software is designed independently in the visual studio development environment, based on the Microsoft Foundation Class Library MFC and open-source computer vision library OpenCV. Rough cloth and light fabric were tested experimentally in the workshop respectively. The experimental results verify the feasibility and effectiveness of the detection system.